On 02/12/2018, 01:14 PM, Jiri Slaby wrote:
> On 02/12/2018, 11:02 AM, Jiri Slaby wrote:
>> Given this happens only on 32bit kernel, I assume some 32bit overflow.
>> But I am unable to see it (yet).
>
> Just to add, a diff of strace in good and bad kernels:
> @@ -655,14 +655,4 @@
> _llseek(3, 4275568640, [4286054400], SEEK_DATA) = 0
> _llseek(3, 4286054400, [4288675840], SEEK_HOLE) = 0
> _llseek(3, 4288675840, [4299161600], SEEK_DATA) = 0
> -_llseek(3, 4299161600, [4301783040], SEEK_HOLE) = 0
> +_llseek(3, 4299161600, [4299161600], SEEK_HOLE) = 2621440
>
> llseek returns a very invalid value when it comes to 0x100400000.
It gets cropped to 0x400000, so:
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -3523,7 +3523,7 @@ static int ext4_iomap_begin(struct inode *inode,
loff_t offset, loff_t length,
iomap->flags |= IOMAP_F_DIRTY;
iomap->bdev = inode->i_sb->s_bdev;
iomap->dax_dev = sbi->s_daxdev;
- iomap->offset = first_block << blkbits;
+ iomap->offset = (u64)first_block << blkbits;
iomap->length = (u64)map.m_len << blkbits;
if (ret == 0) {
sounds about right?
> regards,
--
js
suse labs
